Update README.md
Browse files
README.md
CHANGED
|
@@ -1,23 +1,30 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
# Model Card for RL-GRPO-SQL-Model
|
| 2 |
|
| 3 |
## Model Details
|
| 4 |
|
| 5 |
### Model Description
|
| 6 |
-
- **Model type**: Fine-tuned Model with
|
| 7 |
-
- **Training approach**: Reinforcement Learning with GRPO
|
| 8 |
-
- **Task**: SQL generation
|
| 9 |
- **Developed by**: Ali Assi
|
| 10 |
|
| 11 |
## Training Data
|
| 12 |
|
| 13 |
- **Data sources**: Spider train set
|
| 14 |
-
- **Preprocessing**:
|
| 15 |
-
|
| 16 |
-
## Model Performance
|
| 17 |
-
|
| 18 |
-
### Benchmarks
|
| 19 |
-
|
| 20 |
-
Spider test set
|
| 21 |
|
| 22 |
## How to Use
|
| 23 |
|
|
@@ -26,4 +33,30 @@ from transformers import AutoTokenizer, AutoModelForCausalLM
|
|
| 26 |
|
| 27 |
tokenizer = AutoTokenizer.from_pretrained("ALI-USER/rl-grpo-sql-model")
|
| 28 |
model = AutoModelForCausalLM.from_pretrained("ALI-USER/rl-grpo-sql-model")
|
| 29 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
---
|
| 2 |
+
language: en
|
| 3 |
+
tags:
|
| 4 |
+
- sql
|
| 5 |
+
- code-generation
|
| 6 |
+
- reinforcement-learning
|
| 7 |
+
- text-generation
|
| 8 |
+
datasets:
|
| 9 |
+
- spider
|
| 10 |
+
|
| 11 |
+
---
|
| 12 |
+
|
| 13 |
# Model Card for RL-GRPO-SQL-Model
|
| 14 |
|
| 15 |
## Model Details
|
| 16 |
|
| 17 |
### Model Description
|
| 18 |
+
- **Model type**: Fine-tuned Causal Language Model with Reinforcement Learning
|
| 19 |
+
- **Training approach**: Reinforcement Learning with GRPO (Group Relative Policy Optimization)
|
| 20 |
+
- **Task**: SQL generation and understanding
|
| 21 |
- **Developed by**: Ali Assi
|
| 22 |
|
| 23 |
## Training Data
|
| 24 |
|
| 25 |
- **Data sources**: Spider train set
|
| 26 |
+
- **Preprocessing**: Parsing and validation
|
| 27 |
+
- **Languages**: English
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 28 |
|
| 29 |
## How to Use
|
| 30 |
|
|
|
|
| 33 |
|
| 34 |
tokenizer = AutoTokenizer.from_pretrained("ALI-USER/rl-grpo-sql-model")
|
| 35 |
model = AutoModelForCausalLM.from_pretrained("ALI-USER/rl-grpo-sql-model")
|
| 36 |
+
|
| 37 |
+
# Example usage
|
| 38 |
+
prompt = "Generate SQL for: Find all customers with orders over $100"
|
| 39 |
+
inputs = tokenizer(prompt, return_tensors="pt")
|
| 40 |
+
outputs = model.generate(**inputs, max_length=512)
|
| 41 |
+
print(tokenizer.decode(outputs[0]))
|
| 42 |
+
```
|
| 43 |
+
|
| 44 |
+
## Limitations
|
| 45 |
+
|
| 46 |
+
- Model performance may vary depending on database schema complexity
|
| 47 |
+
|
| 48 |
+
## Ethical Considerations
|
| 49 |
+
|
| 50 |
+
- May generate SQL queries that are inefficient or unsafe if not properly validated
|
| 51 |
+
- Should be used with query validation before execution
|
| 52 |
+
|
| 53 |
+
## Intended Uses
|
| 54 |
+
|
| 55 |
+
**Primary use cases:**
|
| 56 |
+
- Natural language to SQL translation
|
| 57 |
+
- SQL code generation assistance
|
| 58 |
+
- Educational purposes for SQL understanding
|
| 59 |
+
|
| 60 |
+
**Out-of-scope uses:**
|
| 61 |
+
- Direct production deployment without query validation
|
| 62 |
+
- Non-English language queries (not trained for this)
|